| Affiliation: | (1) Department of Bio-engineering, University of Washington, Seattle, Washington, USA;(2) Central Maine Medical Center, Central Maine Heart & Vascular Institute, Lewiston, Maine, USA;(3) Department of Mechanical Engineering, University of Washington, Seattle, Washington, USA;(4) Department of Biomedical Engineering, University of Minnesota, Minneapolis, Minnesota, USA |
| Abstract: | Automatic acoustic classification and diagnosis of mitral valve disease remain outstanding biomedical problems. Although considerable attention has been given to the evolution of signal processing techniques, the mechanics of the first heart sound generation has been largely overlooked. In this study, the haemodynamic determinants of the first heart sound were examined in a computational model. Specifically, the relationship of the transvalvular pressure and its maximum derivative to the time-frequency content of the acoustic pressure was examined. To model the transient vibrations of the mitral valve apparatus bathed in a blood medium, a dynamic, non-linear, fluid-coupled finite element model of the mitral valve leaflets and chordae tendinae was constructed. It was found that the root mean squared (RMS) acoustic pressure varied linearly (r2=0.99) from 0.010 to 0.259 mm Hg, following an increase in maximum dP/dt from 415 to 12470 mm Hg s−1. Over that same range, peak frequency varied non-linearly from 59.6 to 88.1 Hz. An increase in left-ventricular pressure at coaptation from 22.5 to 58.5 mm Hg resulted in a linear (r2=0.91) rise in RMS acoustic pressure from 0.017 to 1.41 mm Hg. This rise in transmitral pressure was accompanied by a non-linear rise in peak frequency from 63.5 to 74.1 Hz. The relationship between the transvalvular pressure and its derivative and the time-frequency content of the first heart sound has been examined comprehensively in a computational model for the first time. Results suggest that classification schemes should embed both of these variables for more accurate classification. |
